Abstract: The complexity of the dilemmas we face on an organizational, societal and global scale forces us into sensemaking activity. We need tools for expressing and contesting perspectives flexible enough for real time use in meetings, structured enough to help manage longer term memory, and powerful enough to filter the complexity of extended deliberation and debate on an organizational or global scale. This has been the motivation for a pro- gramme of basic and applied action research into Hypermedia Discourse, which draws on research in hypertext, information visualization, argumenta- tion, modelling, and meeting facilitation. This paper proposes that this strand of work shares a key principle behind the Pragmatic Web concept, namely, the need to take seriously diverse perspectives and the processes of meaning negotiation. Moreover, it is argued that the hypermedia discourse tools de- scribed instantiate this principle in practical tools which permit end-user control over modelling approaches in the absence of consensus. p. 3

There is a particular need to provide languages for communities to express and contest perspectives in principled ways, using tools that are flexible enough to mediate, capture and add value to discourse between stakeholders, yet introduce sufficient structure to provide computational services without straitjacketing the discourse. This has motivated a research programme focused on co-evolving the semantics, user interfaces, technical infrastructure, and human work practices to embed such tools in highly pressured, real time sensemaking scenarios, face-to- face and over the internet, as well as in scenarios of extended, asynchronous dis- course lasting from a few days to months or years (projects; research discourse). p. 3

This paper seeks to clarify the intellectual and technical threads with which we work, in the hope that they add some distinctive colour and texture. In a nutshell, we seek to provide practical tools for discourse communities to add their own in- terpretive layer over whatever media, syntax and predefined semantics they may be p. 3

The world, indeed our lives, make sense to the extent that we can sustain a coher- ent narrative about who we are and why we matter. If the story fragments, our identity crumbles if we cannot re-integrate it into our narrative. When we are con- fronted by breaches in normality, Karl Weick draws our attention to sensemaking as literally “the making of sense”: sharing interpretations using different represen- tations of the situation. He proposes that: p. 5

“Sensemaking is about such things as placement of items into frame- works, comprehending, redressing surprise, constructing meaning, in- teracting in pursuit of mutual understanding, and patterning.” [We95] p.6 p. 5

Weick’s concern is to characterise what people do in socially complex situations, when confronted by incomplete evidence and competing interpretations : p. 5

“The point we want to make here is that sensemaking is about plausi- bility, coherence, and reasonableness. Sensemaking is about accounts that are socially acceptable and credible. […] It would be nice if these accounts were also accurate. But in an equivocal, postmodern world, infused with the politics of interpretation and conflicting interests and inhabited by people with multiple shifting identities, an obsession with p. 5

accuracy seems fruitless, and not of much practical help, either.” [We95], p.61 p. 6

In other words, when there is uncertainty, what else is there to do but through dis- course, construct a narrative to fill in the gaps? p. 6

Discourse modelling is at once both useful and limited. It is limited in the sense that, like any model, it captures only key features of the world’s richness, in our p. 6

case, the richness of textual prose and verbal discourse.1 However – if done appro- priately – stripping out detail to focus on underlying structure can yield cognitive, computational and theoretical benefits: p. 7

• Cognitive: a well designed external representation exploits the human per- ceptual and cognitive system to direct attention to relevant information;
• Computational: a formal model also provides machines with structure to reason with;
• Theoretical: the removal of detail may assist in identifying generalisable patterns across diverse contexts (see discussion of Cognitive Coherence Re- lations later). p. 7

The function of a medium is to make it possible for people to express, and work with, structure. Sensemaking calls for a particular kind of discourse, expressed through one or more media. Hypermedia can be thought of as the craft, art, science and engineering of managing structure, specifically, relationships, making it the primary discourse modelling medium for several reasons: p. 7

• Modelling discourse relations: an utterance only has meaning in a context, that is, when juxtaposed with others before and after it, and in relation to other possible utterances that make its selection is significant.
• Expressing different perspectives on a conceptual space: diverse stakeholders are usually needed to define and resolve wicked problems, so support tools need to provide support for modelling flexibly, to show agree- ments and differences between viewpoints.
• Supporting the incremental formalization of ideas as understanding de- velops, so that patterns can be captured using representations that are intui- tive, fast in real time usage scenarios, and expressive enough to enable com- putational support.
• Rendering structural visualizations to assist users in grasping complex in- terconnections between ideas and information.
• Connecting heterogeneous content: the content that stakeholders refer to during sensemaking can range from media fragments which offer little or no obvious structure, to material sufficiently structured to support forms of ma- chine reasoning; similarly, relationships may range from associations ex- pressed spatially or as untyped links, to being formally grounded in a known semantic schema. p. 7

4.1. Key characteristics p. 8

• A discourse ontology: a set of explicit constructs that express a subset of the richness of human verbal or written communication.
• One or more notations: symbol system(s) for rendering the ontology. For instance, IBIS can be rendered as a textual outline, and as a directed graph flowing from left to right, or from top to bottom. Each has different affor- dances which can complement each other as coupled visualizations.
• An intuitive user interface: these tools are intended for knowledge workers in diverse sectors of society, not only for modellers with research interests. The notations are therefore just part of designing the overall cognitive and aesthetic experience of working with the tool.
• Computational services: the above come together as augmentation of hu- man capability through software implementation. For instance, “services” would include more efficient capture, interpretation, sharing, retrieval, dis- covery and integration of discourse modelled in the ‘knowledge repository’. Interoperability not only with other relevant tools, but also compatibility with existing work practices will contribute to the overall service augmentation.
• Literacy and fluency: The tool’s functionality is only part of the story, however. We must also examine the capabilities assumed on the part of the user, which we will do under the heading of literacy, the ability to read and write ideas in the new medium in a manner appropriate to the context, ideally moving towards fluency. p. 8

Two discourse model- ling methodologies have developed around the capabilities of Compendium. Dia- logue Mapping is a set of skills developed by Conklin [Co05] for mapping IBIS structures in real time during a meeting in order to support the analysis of wicked problems, as defined by Rittel. p. 9

Conversational Modelling [Se99] incorporates and extends Dialogue Mapping by deriving Issues from a modelling methodology (or for instance, an organizational procedure/best practice). Issue nodes can be saved as reusable issue-template struc- tures to seed different kinds of discussions. p. 9

From a more formal knowledge representation perspective, we represent semantics using a variety of conventions. In a NASA field trial (Figure 3), science metadata was represented using templates which look like visual forms, with each Issue inviting the team to answer (or if necessary debate) the values of the ‘slots’. An issue-template such as this provides a user-friendly way to engage in participatory modelling which permits argumentation if necessary, and results in a set of seman- tic assertions amenable to automated analysis (data entry into a simulation engine in this case). Each Issue in fact embodies the relational semantic connecting its answer to the entity represented by the containing map. However, rather than ask the team to complete sets of semantic triples, they are offered a set of question mark icons to which they need to link lightbulb icons. p. 11

Selvin [Se05] has begun to explore the nature of fluency in what he terms Participatory Hypermedia Construction. Detailed analysis of screen recordings is providing an account of the representational moves that Com- pendium mappers make, and the different roles they can play in meetings. p. 13

ClaiMaker: semantic scholarly publishing and annotation p. 13

ClaiMaker was conceived from the start as a distributed Web application. Its design rationale is the need for representational infrastructure to evolve the current prose document and associated practices for publishing and contesting research results and – equally significant – authors’ interpretations of their significance. p. 13

we are of course dealing with the difficult issue of computational support for an intrinsically pragmatic process, by which a discourse community (in this case, research peers) negotiates what some reported facts should be taken to mean. p. 13

We detail elsewhere [UB06] the design and evaluation of ClaiMaker and the asso- ciated suite of tools for authoring (ClaiMapper) and querying (ClaimFinder) re- search claims and argumentation. These are less mature than Compendium, proof of concept research tools which are not yet publicly available. Space precludes as detailed a treatment as Compendium, but ClaiMaker’s ‘hypermedia discourse pro- file’ below conveys the essence of the approach: p. 13

• Discourse ontology: a two-layer relational taxonomy which provides base relational classes in which ‘dialects’ from different discourse communities are grounded (Figure 4).
• Notation: a conceptual graph of claims that can be visualized using different schemes to show discourse connections between concepts annotated onto the literature.
• User interface: we have investigated a variety of interaction paradigms for annotation tools, in order to help untrained users create semantic annotations [UB06].
• Computational services: the use of a richer discourse scheme than IBIS en- ables us to offer more powerful services. For instance, the semantic citation maps can be filtered in response to queries such as, What documents report data that challenges this author’s hypothesis?What is the lineage of this con- cept: the key ideas on which this work builds? (Figure 5) p. 13

• Literacy and fluency: Being less mature than Compendium, we do not yet have a large enough user community to provide a good description of what it means to read and write such argumentative networks, particularly beyond initial learning. Empirical studies have demonstrated that untrained users can construct and query claim networks [SBM05][UB06]. p. 14

Knowledge modelling from a linguistic, semiotic perspective. While there is substantial activity investigating upper level ontologies of conceptual primitives (entities, time, etc), our focus on discourse has led to work on the relational primi- tives from which domain-specific relational semantics can be derived. We adopt a relation-centric approach to reflect the process of semiosis by which meaning de- rives from, and shifts with, the context of use. This draws on linguistic research into Cognitive Coherence Relations to map between relational grammars across fields and modalities as diverse as scholarly discourse, cinema and text [Ma05][MB06] p. 15

[Bo94] Boden, D., The Business of Talk. 1994, Polity: Cambridge. [BT95] Boland, R.J.J. and R.V. Tenkasi, Perspective Making and Perspective Taking in Communities of Knowing. Organization Science, 1995. 6 (4): p. 350-372. [Bu94] Buckingham Shum, S. and N. Hammond, Argumentation-Based Design Ration- ale: What Use at What Cost? Int. J. Human-Computer Studies, 1994. 40 (4): p. 603-652. [BSD06] Buckingham Shum, S., et al. Memetic: An Infrastructure for Meeting Memory. in Proc. 7th Int. Conf. on the Design of Cooperative Systems. 2006 of Conf.. Carry-le-Rouet. [BSS06] Buckingham Shum, S.J., et al., Hypermedia Support for Argumentation-Based Rationale: 15 Years on from gIBIS and QOC, in Rationale Management in Software Engineering, A. Dutoit, et al., Editors. 2006, Springer-Verlag: Berlin. p. 111-132. [BMS02] Buckingham Shum, S., Motta, E. and Domingue, J. (2002). Augmenting Design Deliberation with Compendium: The Case of Collaborative Ontology Design. in Workshop on Facilitating Hypertext-Augmented Collaborative Modelling, ACM Hypertext Conference. [PrePrint: http://cognexus.org/ht02]. [CSA05] Clancey, W.J., et al. Automating CapCom Using Mobile Agents and Robotic Assistants. 1st Space Exploration Conf.. 2005 of Conf.. Orlando, FL: http://eprints.aktors.org/375 [Co05] Conklin, J., Dialogue Mapping: Building Shared Understanding of Wicked Problems. 2005, Chichester: Wiley. [CB88] Conklin, J. and M.L. Begeman, gIBIS: A Hypertext Tool for Exploratory Policy Discussion. ACM Transactions on Office Information Systems, 1988. 4 (6): p. 303-331. [DA06] de Moor, A. and M. Aakhus, Argumentation Support: From Technologies to Tools. Communications of the ACM, 2006. 49 (3): p. 93-98. [De05] de Moor, A., Patterns for the Pragmatic Web, in International Conference on Conceptual Structures, F. Dau, M.-L. Mugnier, and G. Stumme, Editors. 2005, Springer, Lecture Notes in Computer Science (Vol. 3596). p. 1–18. p. 17

[En63] Engelbart, D.C., A Conceptual Framework for the Augmentation of Man’s Intellect, in Vistas in Information Handling, P. Howerton and Weeks, Editors. 1963, Spartan Books: Washington, DC: London. p. 1-29. [LW91] Lave, J. and E. Wenger, Situated Learning: Legitimate Peripheral Participation. 1991, Cambridge: Cambridge University Press. [MBY91] MacLean, A., et al., Questions, Options, and Criteria: Elements of Design Space Analysis. Human-Computer Interaction, 1991. 6 (3 & 4): p. 201-250. [Ma05] Mancini, C., Cinematic Hypertext. Investigating a New Paradigm. Frontiers in Artificial Intelligence and Applications. 2005, Amsterdam: IOS Press. [MB06] Mancini, C. and S. Buckingham Shum, Modelling Discourse in Contested Do- mains: A Semiotic and Cognitive Framework. In Press. Int. J. Human Computer Studies. [RS03] Repenning, A. and J. Sullivan. The Pragmatic Web: Agent-Based Multimodal Web Interaction with no Browser in Sight. in Proc. INTERACT'03. 2003 of Con- ference: IOS Press, IFIP, pp. 212-219. [Ri72] Rittel, H.W.J., Second Generation Design Methods. Interview in: Design Meth- ods Group 5th Anniversary Report: DMG Occasional Paper, 1972. 1: p. 5-10. [SDD06] Schoop, M., A. De Moor, and J. Dietz, The Pragmatic Web: A Manifesto. Com- munications of the ACM, 2006. 49 (5): p. 75-76. [PrePrint: http://doi.acm.org/10.1145/1125944.1125979]. [Se99] Selvin, A., Supporting Collaborative Analysis and Design with Hypertext Func- tionality. Journal of Digital Information, 1999. 1 (4). [Se05] Selvin, A., Aesthetic and Ethical Implications of Participatory Hypermedia Practice. 2006, Technical Report KMI-05-17, Knowledge Media Institute, Open University. [SB02] Selvin, A.M. and S.J. Buckingham Shum, Rapid Knowledge Construction: A Case Study in Corporate Contingency Planning Using Collaborative Hyperme- dia. Knowledge and Process Management, 2002. 9 (2): p. 119-128. [SBM05] Sereno, B., S. Buckingham Shum, and E. Motta. ClaimSpotter: An Environment to Support Sensemaking with Knowledge Triples. 10th ACM Conf. on Intelli- gent User Interfaces. 2005 of Conf.. San Diego: ACM Press: NY. [SDD06] Schoop, M., A. De Moor, and J. Dietz, The Pragmatic Web: A Manifesto. Com- munications of the ACM, 2006. 49 (5): p. 75-76. [Si02] Singh, M.P., The Pragmatic Web. IEEE Internet Computing, 2002 (May/June): p. 4-5. [TBD05] Tate, A., et al., Co-OPR: Design and Evaluation of Collaborative Sensemaking and Planning Tools for Personnel Recovery. 2006, Technical Report KMI-06-07, Knowledge Media Institute, Open University. [UB06] Uren, V., et al., Sensemaking Tools for Understanding Research Literatures: Design, Implementation and User Evaluation. Int. J. Human Computer Studies, 2006. 64 (5): p. 420-445. [We95] Weick, K.E., Sensemaking in Organizations. 1995, Thousand Oaks, CA: Sage Publications. p. 18